[AstroPy] Deployment and packaging
luigi at lambrate.inaf.it
Thu Jun 16 08:26:38 EDT 2011
Il 06/16/11 13:59, Tom Aldcroft ha scritto:
> On Thu, Jun 16, 2011 at 7:01 AM, Éric Depagne<eric at depagne.org> wrote:
>>> Is Linux, then, the main problem as far as "easy Python
>>> [+numpy/scipy/matplotlib] installations" go?
>> To my knowledge of various distributions, installing python/numpy/scipy is
>> very easy when running linux, it's just a matter of asking your package
>> installer to do so, and here it goes.
>> Of course, one has to know that it's done this way, but I suspect that it's
>> not the biggest hurdle (maybe I'm wrong, I do not know).
> This requires root access. At CfA all the system-managed linux
> machines (hundreds) run CentOS-5, which includes Python 2.4 as the
> native python. The RPM installs for numpy / scipy will be ancient
> versions. Using CentOS-5 (as I am forced to do) is a challenge since
> the libraries are likewise ancient. Getting PyGTK to build from
> source was a major project in archaeology, trying to guess which
> versions of the gtk dependencies would build.
Gosh, PyGTK remembers me pangocairo and family... a nightmare!
My experience is that the most tricky libraries to be installed are just
those related with a graphical interface system (we used Tk and GTK, but
now we would prefer Qt): installing python, numpy and matplotlib is a
minor problem (concerning scipy I'm not so sure it is so easy though).
> The upshot is that to provide a universally applicable solution
> requires a non-root installable distribution.
> I think everyone agrees that EPD itself is not acceptable because of
> license restrictions. Probably 2/3 of people at CfA would not qualify
> under the academic license, myself included.
It is not acceptable, but certainly comfortable.
More information about the AstroPy